Thanks for the links. As I previously commented, we are relying on this unicode standard list.
Unfortunately we don’t have the bandwidth right now to change this, we have been working for months on the new accents and language strategy and we won’t be able to advance on this field until that’s resolved and implemented (hopefully mid-end this year).
Note this strategy includes a way to have Romansh as a dataset language and capture the different Romansh sound variations, which will allow us to capture people speaking in all the variations you mentioned.
Thanks for your understanding.